Emotion in lexicon and grammar: lexical-constructional interface of Mandarin emotional predicates

1.1 The issues

The eventive predication will normally correspond to the verbal passive (with by-PP), indicating a semantic distinction from the adjectival passive. While some of these issues have been dealt with in previous works, few studies have given a comprehensive analysis of the full range of grammatical distinctions in the emotional lexicon.

1.2 The scope and goal

With the aim to investigate EPs in Mandarin Chinese, a major non-inflectional language outside the Indo-European family, this study will further elaborate on the abovementioned distinctions and inquire about the possible range of lexical-constructional associations that are semantically distinct and grammatically realized in the lexicon of emotion.

Specifically, this study attempts to probe into the semantic correlates of the possible grammatical variations to see if the contrasts manifested in English are universal and essential to EPs. It takes on three main questions: (1) How is emotion conceptualized and lexicalized in a non-European language such as Mandarin? (2) How do emotional predicates differ from each other? That is, what are the lexical-constructional variations displayed among EPs? (3) How can the study of EPs shed light on typological and theoretical issues in verbal semantics?

In its attempt to answer the above questions, the study will provide a cognitive semantic account of the range of lexicalization patterns attested in the Mandarin lexicon. The ultimate goal of the study is to identify reliable semantic-to-syntactic criteria for establishing the subclasses of the Mandarin EP inventory for further language-specific representation or cross-linguistic comparison.

1.3 Summary of findings

A closer examination of the major works on the lexical semantic distinctions of EPs reveals that besides the two commonly recognized participant roles, Experiencer and Stimulus, another semantically distinct and non-decomposable role, Affector, is also prominent in emotional predication, as it profiles a higher degree of volitional impact. Affector can be defined in relation to the notion of affectedness that is taken to be a scale of change the theme participant undergoes (Beavers 2011, 2013; Tenny 1987, 1992; Kenny 1963). Different from the non-sentient Stimulus, an Affector volitionally instigates an internal change on an Affectee in a more dynamic and eventive manner. With the postulation of Affector, a three-way distinction of lexicalization patterns is in place that is syntactically attested in Mandarin. Namely, EPs may lexicalize either an Experiencer, a Stimulus, or an Affector as the subject, with different implications of eventivity and degree of affectedness. The distinction of three types of subject roles correspond nicely to the three-way case distinction that surfaces in some Indo-European languages that have three different cases for EP subjects (see discussion in Section 2.2). It also helps to account for the stative vs. eventive distinction as mentioned above and illustrated in 5. The proposed three-role scheme thus provides a sound basis for lexical semantic categorization as well as cross-linguistic comparison. While verbs can be categorized into different subtypes based on subject role selection, languages may vary in terms of the predominant and preferred pattern of lexicalization as a particular role may be most frequently chosen in the emotion lexicon.

In view of the saliency of subject roles in determining the subtypes of EPs, the study further probes into the range of form-meaning mapping principles realized in polysemous relations as well as the grammatical means typically drawn upon to encode a subject role shift. It is found that the Chinese emotion lexicon is unique and differs from the known European languages in two respects. First, Chinese lacks readily lexicalized Stimulus-subject verbs, i.e., equivalents to English verbs such as interest, please, and frighten. Instead, a range of constructional templates are unutilized to derive Stimulus-subject EPs that are semi-lexicalized and morphologically-open. Secondly, Mandarin EPs display a unique spectrum of polysemous relations in that the same verb form may be associated with multiple subject roles and grammatical functions, demonstrating a heterogeneous range of form-meaning mismatches. It is interesting to note that the Chinese way of coining Stimulus-subject EPs via constructional modes and its lexical association with multiple usages both demonstrate that lexical and constructional entities are meaning-bearing units that constitute a continuum of form-meaning association along the same dimension, an important observation that is potentially in line with the theoretical premises of Construction Grammar (Goldberg 1995, 2005).

1.4 The proposal and organization

Based on the findings, the study eventually makes three main proposals: (1) There are three different subject roles (Experiencer, Stimulus, and Affector) that need to be distinguished for the classification of EPs in Mandarin; (2) Different from English, Mandarin tends to prefer lexicalizing Experiencer or Affector as subject as it lacks fully lexicalized Stimulus-subject verbs³; (3) Given the three-way distinction, a language may be Experiencer-prominent, Stimulus-prominent, or Affector-prominent, depending on the relevant factors of constructional unmarkedness, lexical status, and distributional frequency. Languages can then be compared in terms of their predominant lexicalization patterns⁴. By deciphering the collo-constructional variations across different lexical classes, the study ultimately shows that lexicalization works hand-in-hand with constructionalization in shaping the lexicon of emotion.

Following the Introduction, Section 2 reviews a series of studies on lexicalization patterns and semantic distinctions of EPs in English and other languages. Section 3 provides a preliminary account of Mandarin EPs and explores the lexical-constructional interactions characteristic of the Mandarin emotion lexicon. Section 4 discusses the form-meaning mismatches displayed in some particular classes of EPs. Section 5 outlines the preliminary typological distinction. Section 6 draws on theoretical implications and concludes the study. The analyses are mainly based on observations of corpus data from Sinica Balanced Corpus 5.0 of Modern Mandarin, supplemented with data from Chinese Gigaword Corpus⁵. Whenever necessary, contrastive skeleton examples may be given for the sake of clarity.

2.1 Two-way distinction: subject selection

Apparently, with a Stimulus-subject verb like frighten, English systematically utilizes grammatical derivations to allow an Experiencer to take the subject position. This pattern appears to be fairly productive, compared to the more limited pattern in deriving a Stimulus subject with a lexically Experiencer-subject verb like “fear”. This is why Talmy further argues (2000: 98) that “while possibly all languages have some verbs of each valence type, they differ as to which type predominates. In this respect, English seems to favor lexicalizing the Stimulus as subject.…The bulk of its vocabulary items for affect focus on the Stimulus.” In contrast to English, Atsugewi (a Native American language) was mentioned as having verb roots that exclusively take an Experiencer-subject. To express a Stimulus-subject in this language, a suffix –ahẃ has to be added to the verb root.

Talmy’s dichotomy of emotional valence in terms of subject selection serves as a preliminary and convenient scheme to categorize the emotional lexicon. However, as Talmy (2000: 99) cautions, the boundaries of the “affect” category may be too “encompassive or misdrawn” and there may be “smaller categories following more natural divisions that reveal more about semantic organization.” And indeed, EPs are semantically heterogeneous and more complex distinctions are proposed in other works as summarized below.

2.2 Three-way distinction: case marking

As Filip further mentioned, a similar tripartite division in case marking can be found in other Indo-European languages, including French (Legendre 1989), Italian (Perlmutter 1984; Belletti and Rizzi 1988), Dutch (Zaenen 1988), Russian (Holloway-King 1993), Bulgarian (Slabakova 1994), and in South Asian languages (cf. Verma and Mohanan 1990). The three lexical classes are further accounted as displaying different clusters of semantic features along the Proto-Agent vs. Proto-Patient paradigm (Dowty 1991). Two of the salient Proto-Agent properties, “sentience” and “volition,” are taken to be the crucial motivation for the nominative case marking, while the accusative and dative case markings imply a lack of control or a low degree of control on the part of the Experiencer, displaying more Proto-Patient characters as being affected under the direct or indirect force of an external Stimulus.

The three-way distinction proposed indicates the presence of a separate thematic role from the traditional notion of Stimulus. A more dynamic and agentive role is apparently involved in the inchoative or eventive version. To highlight its affective role, this type of subject can be called Affector, which instigates a change on the object, the Affectee, which undergoes the change, as marked with accusative case in Czech. The Affector plays a similar role as what is termed “Effector” in Van Valin (2005), a more fundamental notion than Agent that underpins the basic properties of a volitional and acting instigator. More detailed discussion will be given in Section 3.

2.3 Four-way distinction: argument expression

It is noted that Levin chose the term “cause,” instead of stimulus, in describing the subject of the amuse group of verbs by saying “they are transitive verbs … whose subject is the cause of the change in psychological state” (1993: 191). She then indicated, following Grimshaw (1990), that some of these verbs, such as amuse, allow the subject argument to receive an agentive interpretation, while others, such as concern, do not. This distinction, as further suggested by Levin, could be the basis for further subdivision of this group of verbs. And indeed, as we already seen in the Czech examples, the agentive interpretation of the subject suggests a semantically distinct “causer” role that is termed Affector in this paper.

2.4 Five-way distinction: morphological variants

This meaning distinction interacts with grammatical forms and underlines the argument structure of Experiencer-subject adjectives, since inherent feelings will not require the presence of a Stimulus as shown in 10a, but Stimulus is required and cannot be left out in 10b⁹. As will be clear in the next section, this lexical semantic distinction also bears grammatical consequences in Mandarin and it is significant in fine-tuning Mandarin near-synonyms.

Secondly, the Exp-V-Stim frame in 10c is considered to be conforming to the same predication template with the adjectival use as it “incorporates” the predication function BE. In other words, the English verbal use is semantically similar to the adjectival use. This analysis is of particular interest to the Mandarin lexicon, since Mandarin does not draw a clear line, morphologically, syntactically, and semantically, between stative verbs and adjectives. In view of Jackendoff’s analysis, stative verbs and adjectives share the same semantic function BE and thus help to justify the null distinction between verbal vs. adjectival predicates in Mandarin.

Thirdly, the adjectival frames, Exp-Adj-Stim in 10b and Stim-Adj-(Exp) in 10e, are taken as describing the same situation, since the adjective with a Stimulus subject (e.g., Golf is interesting to Bob) is derived as a paraphrase of the Experiencer-subject adjective (Bob is interested in Golf), with the so-called lambda-abstraction: Golf is such that Bob is interested in it. The paraphrase is formally achieved with the effect of marking the Stimulus as prominent (see Jackendoff 2007: 228 for details). Interestingly, the suggested semantic connection between Experiencer-subject and Stimulus-subject frames is also realized in Mandarin. The same verb can be used for both purposes with a constructional shift.

Finally, the transitive Stim-Verb-Exp frame (This bores me.) in 10d is treated as conceptually synonymous with the adjectival frame (This is boring (to me).) in 10e, as they are both causative in nature (cause X to BE). Their conceptual similarity in causality is grammatically realized in Mandarin as the causative pattern is commonly used to predicate a Stimulus-subject in either transitive or intransitive use, as will be detailed in Section 3.1.

In distinguishing the four variants, Jackendoff takes “causative” as having an agent subject, which may or may not be identical with a Stimulus. The agent subject serves as an affecting causer that impacts the Experiencer and makes it more like a patient. This is in line with the role hierarchy proposed in Pesetsky (1995): Causer > Experiencer > Target (or subject matter). Pesetsky’s notion of Causer can be either an agentive or presumably non-agentive Stimulus subject. The more agentivity is perceived on the Causer, the more affectedness is rendered on the Experiencer, who resembles an undergoer of impact without much control¹⁰. Judged by the varied degrees of impact on the Experiencer, it is clear that the agentive subject can be distinguished and separately considered from the non-sentient Stimulus. This supports the postulation of an agentive subject role (the Affector), accompanied with a patient-like object role (the Affectee). Again, the functional correlation observed in English is grammatically confirmed in Mandarin.

What is striking here is that most of the semantic implications postulated in Jackendoff (2007) may find grammatical evidence in Mandarin. It will be shown in subsequent discussions that the distinction between Experiencer-subject EPs (inherent vs. directed feelings as in 11) can be further elaborated with studies on Mandarin near-synonym sets of Experiencer-subject EPs since they are abundant and semantically fine-grained in Mandarin. And, it will be clear that Stimulus-subject predication in Mandarin essentially involves finer distinctions of causal relation, since it is overtly expressed with a marked causative construction (Stimulus as causer).

2.5 Distinction of near-synonyms: causal and constructional variation

A number of pioneering works on Mandarin emotion lexicon looked specifically into the syntax-semantics interface manifested in Experiencer-subject EPs, with a focus on commonly recognized near-synonym sets (1999; Chang et al. 2000; Liu 2002). These works aim to discover the fine-grained semantic distinctions from a corpus-based approach. Among them, Tsai et al. 蔡美智等 (1999) examined the frequently used pair of verbs, 高興 gaoxing “be glad, pleased” and 快樂 kuaile “be happy, content”. It is found that 高興 gaoxing displays a higher frequency in predicative use (vs. nominal use), eventive adverbials, and causal complements. Based on distributional differences in nominalization, adjectival/adverbial modification, and sentential complement, the study proposed that the two verbs differ with a semantic distinction in inchoative vs. homogeneous state (ibid.: 449-453). The distinction is further decomposed into two semantic features: change of state and control. The verb 高興 gaoxing represents an inchoative state with higher degree of experiencer control and is thus lexically specified with the features <+change of state, +control>, while 快樂 kuaile, represents a homogeneous state with less volitional control and is thus characterized as <−change of state, −control>.

Following up on the above study, Chang et al. (2000) looked at more sets of Experiencer-subject EPs and proposed a morphological account for the systematic variation between inchoative vs. homogeneous state verbs. Seven semantic fields of emotional sentience are distinguished, including happiness, worry, fear, anger, regret, sadness, and depression. For each field, two representative lemmas are examined as a contrastive pair. Based on distributional criteria, verbs in each field are divided into two semantic types: (1) type A (inchoative state): 高興 gaoxing “become glad/pleased,” 難過 nanguo “become sorrowful,” 後悔 houhui “regret,” 傷心 shangxin “get heart-broken,” 生氣 shengqi “get angry,” 害怕 haipa “get scared,” 擔心 danxin “get worried”; (2) type B (homogeneous state): 快樂 kuaile “be happy,” 痛苦 tongku, “be painful,” 遺憾 yihan “be sorry,” 悲傷 beishang “be sad,” 憤怒 fennu “be full of anger,” 恐懼 kongju kong “fear,” 煩惱 fannao “be perplexed”.

The semantic types with their paradigmatic differences in grammatical distribution are further attributed to morphological differences: type A verbs are mostly non-VV compounds (e.g., Manner-V for 高興 gao-xing “high-excite” and 難過 nan-guo “difficult-pass”) and type B verbs are VV compounds (e.g., 快樂 kuai-le “cheer-merry,” 痛苦 tong-ku “ache-suffer”). The VV compounds combine two synonyms (or antonyms) to represent a functionally uniform kind of emotion, which is semantically more homogeneous and time-stable. On the other hand, non-VV compounds may involve a skewed combination of functionally distinct elements (Manner + Verb or Verb + Goal) and are prone to predicate a change of state or inchoative event.

The Stimulus-as-Cause construction highlights an external cause that is important in distinguishing the meanings of the two verbs. Liu (ibid.) further distinguishes two types of caused events, adopting the analysis of verbs of sound in Atkins et al. (1996). Similar to sound emissions, emotional experiences can be externally or internally caused, which also bears pragmatic implications. The socially more acceptable verb 羨慕 xianmu “envy” is taken to be externally caused as it has a higher percentage of verbal use and tends to collocate with an externally describable cause, which may serve as a social justification of the emotion. In contrast, the less acceptable counterpart 忌妒 jidu “be jealous of” is internally-caused, as it is has a higher percentage of nominal, non-causal use and if a cause is ever present, it is typically an inner, non-describable cause, such as 心 xin in 心生忌妒 xin-sheng-jidu “jealousy from within”. The absent or incommunicable cause makes it hard to be socially justifiable.

Liu (2002)’s proposal of internally vs. externally caused emotions may provide a principled account to integrate the distinction between inchoative vs. homogeneous state in Tsai et al. 蔡美智等 (1999) and the division of type A vs. type B verbs in Chang et al. (2000). Inchoative states or type A verbs are externally caused, as they involve a higher degree of volitional control with verbal use and more overt mention of an externally present cause. On the other hand, homogeneous states or type B verbs are internally caused, involving a higher percentage of nominal use with less volitional control and less mention of a describable cause.

The proposed analysis also goes nicely with the distinction of inherent vs. directed feelings (bored vs. interested), proposed in Jackendoff (2007) and illustrated in 11 above. As Jackendoff asserts, inherent feelings are “pure emotions” that are independent of the external surroundings and thus may not require an external cause (i.e., internally caused in Liu (2002)’s terms), while directed feelings require the mention of an external stimulus (i.e., externally caused). This fine-grained distinction is applicable to most Experiencer-subject predicates, transitive or intransitive.

Based on findings on cultural universals (Ekman and Davidson 1994), Jackendoff further states that the difference between inherent vs. directed feelings is psychologically founded and “does not appear to have anything to do with language” (Jackendoff 2007:225). It is clear that this semantic distinction may be universal and cross-linguistically applicable, as also evidenced in Mandarin.

What follows is a close examination of the factors in relation to the Mandarin emotion lexicon, which will provide a more complete picture to address the original concern: what is unique in the lexicalization of Mandarin EPs? To what extent and in what way can the observed lexicalization patterns be considered cross-linguistically relevant?

3.1 Preliminaries

The majority of Mandarin EPs are stative in nature as they are compatible with degree evaluation. But a degree marker is not obligatory with stative EPs, if the verbal use is to be stressed¹⁴. Given that there appears to be a functional division between 很hen “quite, fairly,” marking evaluative predication, and the perfective/inchoative marker 了 le, marking eventive predication, EPs can be divided as to their co-occurrence preference with the two markers. Lexical variations can be found with a group of Stimulus-subject verbs that tend to align more with eventive predication, indicating a semantic departure from stativity, which will be further discussed in the next section.

Given that there is no morphological difference between stative verbs and adjectives in Mandarin, the verbal vs. adjectival contrast lexically coded in English (envy vs. be envious of, as illustrated in 2 above) is formally neutralized and indistinguishable in Mandarin. For example, the abovementioned EP 羨慕 xianmu can be taken as either “envy” or “be envious of,” especially with the presence of a degree marker.

Due to the lack of Stimulus-subject verbs in Mandarin, the causative pattern is utilized as a grammatical strategy to allow the shift of subject roles. What we see here is that while Stimulus-subject predication is lexically encoded in English, it is mostly done at the constructional level in Chinese. This points to an interesting and significant departure in lexicalization patterns, as will be further discussed in the section on subject role shift.

The preliminary introduction shows some unique features of Mandarin EPs. First, degree modification is compatible with Mandarin emotion predication, neutralizing the difference between stative verbs and adjectives. This prepares further discussion of the stative-eventive distinction that may be lexically distinguished in Mandarin. Secondly, Mandarin seems to allow more flexibility in argument expression and one lexical form may be mapped to more than one grammatical function, which leads to a further discussion of the range of form-meaning mapping relations manifested in polysemous EPs. Thirdly, Chinese is apparently more limited in lexicalizing Stimulus-subject verbs and it utilizes constructional means to remedy the missing link, which leads to a further discussion of the lexical-constructional variations characteristic of the Mandarin emotion lexicon.

3.2 Formal marking of the stative vs. eventive distinction

These verbs exemplify the lexicalized predicates that morphologically encode a salient impact or change of state in the form of V-R¹⁹, which is semantically less compatible with the indicator of stativity (很 hen), but prefers the inchoative/perfective aspectual marker 了 le. The V-R compounds stand out as a distinct set of EPs that are semantically eventive with a lexically assured change of state, normally indicating a higher degree of agentivity and affectedness. Although the subject may not always be human and volitional, the transitive event constantly collocates with the aspectual marker 了 le, which is associated with eventive predication and profiles a temporal boundedness. This group of EPs shows that the Mandarin emotion lexicon is sensitive to the stative-eventive distinction as already mentioned and illustrated in 5 above.

This stative vs. eventive semantic distinction may be lexically implicit in English, but it is more explicitly encoded in Mandarin, corresponding to morphological and constructional differentiations. What needs to be noted here is that the lexicalized eventive verbs in the form of V-R must denote a different semantic relation from the traditional Stimulus-to-Experiencer relation.

Relevant to the questions raised in 14, the range of semantic variations illustrated above is manifested with a range of lexical-constructional variations. Such variations are motivated by a gradation from highly stative to highly eventive distinctions in the uses of Mandarin EPs. In the last example with a human subject, it is quite clear that the subject plays a more volitional and instigating role, different from the non-active, non-sentient role of a Stimulus. This role distinction is relative to the extent of affectedness instilled on the theme participant and should be recognized as a lexical semantic distinction. As will be clear in the discussion of the thematic relation involved, the volitional subject may be more appropriately viewed as an Affector, if not a prototypical agent.

3.3 Constructional variation with subject role shift

As noted earlier, the causative construction is productive and constantly drawn upon to coin the missing Stimulus-subject EPs, due to the lack of lexical Stimulus-subject verbs such as please, excite, and interest. The closest equivalents of pleasing, exciting and interesting are semi-lexicalized causatives, derived from the causative template with a generic causee in the form [CAU-person-V], such as 令人高興 ling-ren-gaoxing [CAU-person-happy], for “pleasing”. This impersonal causative pattern behaves like other stative predicates since it can also take a preceding degree marker such as 很 hen ²¹, but it is not yet fully lexicalized (see discussion in Section 3.5). It provides a grammatical means for shifting the subject role from Experiencer to Stimulus while maintaining a semantic link using the same verb.

Given the dual subjecthood, 無聊 wuliao may still be used in the above-mentioned causative pattern as other Experiencer-subject EPs, to highlight a Stimulus-causer, as in 這本書令人無聊 zhe-ben shu ling-ren-wuliao [CAU-person-bored] “The book is boring.”

The abovementioned verbs are semantically and syntactically diverse, challenging the traditional lexical divisions based on semantic roles and argument expressions. What is of particular interest here is that the dual-subject verbs are able to predicate both Experiencer and Stimulus, breaking down the basic line between Experiencer-subject vs. Stimulus-subject verb classification.

The distinction with degree modification in the passive constructions, pertaining also to the stative-eventive distinction, can be compared with the adjectival vs. verbal passive distinction in English (with-pp vs. by-pp), as illustrated in 4 above. In English, the semantic distinction is not clear in the active voice but may be syntactically surfaced in the passive version with different prepositions. The verbal passive (with by-PP) may convey a similar function as the Mandarin eventive passive, which signals a more affective relation between the subject and object. This leads to the postulation of a different set of semantic roles, Affector and Affectee, in the next section.

A finer distinction of the semantic roles of the subject is necessary, as proposed below, to help differentiate the observed variations in the subclasses.

3.4 Distinction of thematic roles: Stimulus-Experiencer vs. Affector-Affectee

The fact that a causative structure is called upon to express a Stimulus-subject relation suggests that the role of a Stimulus is taken to be functionally identical to a Causer. This has been mentioned to confirm what Pesetsky (1995: 56) proposed regarding the hierarchy of assigning thematic roles to subjecthood: Causer > Experiencer > Target/Subject matter. The hierarchy helps to point out the essential role of a Causer in emotional predication. When a Stimulus becomes the subject, its semantic function as a Causer is overtly expressed with the overtly marked causative construction in Mandarin. This syntactic strategy with constructional shift strongly suggests that the relation from Stimulus to Experiencer is fundamentally causal.

In view of the comparison, we see that finer distinctions of affectedness in terms of realization of change (Beavers 2011, 2013) may be both lexically and grammatically differentiated. Examples with 激怒 jinu “infuriate” apparently allow the subject to exercise more control over the directly affected object. The semantic distinction, as mentioned previously, is referred to by Jackendoff (1991:140) as the stative vs. eventive distinction on Stimulus-subject verbs, and noted in Levin (1993: 191) as agentive vs. non-agentive role distinction. Dowty (1991: 580) attributed the inchoative (his term for “eventive”) use to the entailment of the Proto-Patient property in the object, reminiscent of the accusative marking of Experiencer in the Czech data. In Mandarin, there is even a stronger correlation of the semantic distinction with formal differentiations. The highly change-entailing verbs (e.g., 激怒 jinu “infuriate”) are morphologically distinct as V-R compounds and syntactically distinct in taking dynamic aspectual markers (perfective 了 le or progressive 在 zai). They lexically encode a “change of state” that is morphologically attained with the second component in the sequence of V-R, literally combining an active verb 激 ji “stir” and a resultative 怒 nu “angry” (lit. “stir-anger”).

This distributional difference may add to the evidences that point to a finer distinction of semantic relations. The Mandarin grammar makes it clear that the agentive vs. non-agentive subject roles and the stative vs. eventive predications are lexically and constructionally distinct and may constitute distinct subclasses in the lexicon.

Liu and Hu (2013) proposes that although the alternating expressions were traditionally thought to be semantically similar, they involved a form-meaning re-assignment and should be considered two very different constructions that project two different views of excessive emotion, involving two distinct sets of thematic relations. An excessive emotion can be expressed either as an sentient state in the excessive degree construction (EDC) that encodes the typical Experiencer-to-Stimulus relation or as an impactive event in the excessive impact construction (EIC) that highlights the forceful relation between Affector and Affectee. The two constructions are evidenced by distributional asymmetry in participating verbs and temporal collocates. It is found that only DIC may take verbs of perceptual effect or physical action (such as 酸 suan “sour” or 寫 xie “write”) and collocate more with temporally-situated specifiers (such as 剛才 gangcai “just now,” 今天 jintian “today”). The proposed constructional distinction helps to confirm the semantically distinct relation between Affector and Affectee, which finds its extended use in excessive expressions.

3.5 Distinction in lexical status

The intriguing interactions between lexical and constructional patterning will be further explored with regard to form-meaning pairing relations in polysemous predicates in Mandarin.

4.1 Verbs with dual subject roles

The verb 無聊 wuliao “be boring/bored” aligns two different semantic roles (Stimulus or Experiencer) in the same grammatical position (intransitive S). Although contextual information may normally single out one reading, the verb is potentially ambiguous at both the lexical and constructional levels: one single form is mapped with two possible meanings.

The deliberating causative use in 46c is comparable to the active use of annoying in English. As observed in Jackendoff (2007: 234), some causative verbs in English give rise to “causative adjectives” that can be used actively in the progressive aspect, denoting a manner calculated to cause annoyance²⁵, as in Harry is being annoying.

As shown previously, the V-R verbs are highly eventive with more impact of change, which may show the best match with frequency adjunct and 被 bei-passive in profiling the Affector-Affectee relation; verbs that do not fully match with event-individuating frequency and 被 bei-passive are more stative with less impact of change in coding the Stimulus-Experiencer relation: The cross-categorical verbs in the middle may be stative or eventive, showing a wider range of uses.

The above cases demonstrate different types of dual-subject or multi-subject EPs that may alternate between Experiencer vs. Stimulus, or Stimulus vs. Affector subject roles without variations in lexical form, while such alternations are normally accompanied with derivations in an inflectional language such as English. Next, we will look at EPs that lexically specify a single subject role but may be used transitively or intransitively.

4.2 Verbs that demonstrate split intransitivity

What needs to be noted is that the unaccusative verbs such as 滿足 manzu “satisfied/satisfy” that align S and O roles are traditionally viewed as unmarked “lexical causatives” since it lexically encodes a causative (transitive) relation, in a stative or eventive sense, without using an overt causative marker.

Given the above cases, it is not uncommon to find Mandarin EPs with dual or multiple usages (see Appendix for a more exhaustive list). Among all the subclasses, the group of Stimulus-subject, transitive verbs are most likely be used in some other ways. It may be more precise to say that they tend to come from other categories, either from Experiencer-subject or Affector-subject EPs.

4.3 Cross-categorial verbs: an extreme case

The diverse uses of 煩 fan encompass all five possible senses attested in Mandarin emotional predicates. It demonstrates the widest range of semantic extension and lexical flexibility.

4.4 Form-meaning Mismatch and constructional pragmatics

The unusual causative use of the Stimulus-subject verb 麻煩 mafan would typically collocate with the second-person honorific form to express extreme politeness. The example represents a type of form-meaning mismatch that helps to fulfill a pragmatic function, i.e., to express indirectness or super-politeness. The lexical-constructional mismatch is a formally marked strategy for a pragmatically marked expression. The unique usage appears to illustrate an interesting case of constructional pragmatics in that the Stimulus-headed causative construction, with a redundant Stimulus-subject verb, is loaded with pragmatic inference for indirectness and extra-politeness.

In terms of grammatical coding of semantic roles, the verb 麻煩 mafan “trouble/be troublesome,” demonstrates an unergative pattern with a Stimulus subject, aligning S and A positions for the role Stimulus. In contrast, the previously discussed verb 討厭 taoyan “to detest/be detestable,” displays the unaccusative counterpart in aligning S and O positions for Stimulus.

From the above summary, we see that the Mandarin lexical patterns are highly correlated with thematic relations, lexical status, argument realization, and constructional associations. With the above discussions, we are ready to make a preliminary proposal regarding the major lexical-constructional patterns of the emotional lexicon.

5.1 Three-way distinction on subject roles

The example was given to differentiate Agent from Stimulus. Jackendoff made it clear that in such cases, the subject is no longer a Stimulus; it is only an agent²⁷. It happens that Mandarin has gone a step further to encode a clear distinction between the two subject roles with lexical-constructional variations to profile an agentive role, the Affector, which adds a higher volitional force to the underlying causative role of an instigator.

This wisdom is fully realized in Mandarin, since the Mandarin way to obtain a Stimulus-subject expression is overtly causative, with syntactically marked with a causative construction or morphologically marked with a semi-lexicalized causative, both with a causative marker, as adequately illustrated in previous discussions.

5.2 Subclasses of emotional predicates

Some of the verbs may be cross-categorial as discussed and exemplified previously. A detailed list of the representative lemma for each class is given in the Appendix. As the Appendix shows, the Stim-subj predicates are most problematic among the five subclasses. The intransitive ones are mostly syntactically derived and morphologically open, and the transitive ones are mostly cross-categorial. This suggests that the category of Stim-subj verbs may not be as lexically rich and stable as the other categories, a fact that has a typological significance in characterizing the Mandarin emotional lexicon. The five lexical-constructional patterns may be used for a cross-linguistic comparison of the emotion lexicon as languages may vary in their preferred categories.

5.3 Preferred emotional pattern

While the three lexicalization patterns may co-exist in a language, they may vary in their prominence as the preferred pattern. As Talmy (2000) pointed out, there are usually a set of possible lexicalization patterns attested in the world’s languages, but one specific pattern may normally be identified as being “characteristic” of a given language. However, what serves as the criteria for identifying the characteristic pattern may be an empirical issue. In view of the Mandarin data, the predominant pattern may be identified with three criteria: (1) constructional unmarkedness, (2) lexical status, and (3) frequency. The three criteria are characterized respectively as follows:

Potentially, a language may prefer to lexicalize an Affector in the unmarked construction to express emotion in a more dynamic and affective way. This will then give rise to an Affector-prominent language, which awaits to be attested with further investigation.

Lexical status refers to the morphological discreteness and stability of a lexical entry. A fully lexicalized entry will be formally stable and distinct. In Mandarin, the Experiencer-subject predicates exhibit a higher degree of lexicalization as fixed form-meaning pairing entities, while Stimulus-subject verbs are syntactically derived from various structural templates and show a higher degree of morphological compositionality. These predicates remain morphologically open and less fixed in lexical forms.

As can be seen from the italic numbers in the chart, Exp-subject uses are more frequent than Stim-subject uses. In view of the above three criteria, Mandarin is considered to be Experiencer-prominent. It is typologically contrastive to English, a Stimulus-prominent language. It remains questionable to see if any language is Affector-prominent, profiling a more dynamic view of emotional events. Based on the Mandarin data, it may be speculated that Experiencer-prominence may be more common than Stimulus-prominence cross-linguistically, since Stimulus-subject EPs are not readily available in Mandarin. Further cross-linguistic research could help to establish a more substantial typological account²⁸.